智能水位预测服务系统研究  

Research on intelligent water level prediction service system

作　　者：潘明阳[1] 周海南 李增辉 刘乙赛 李超[1] 李昱 PAN Ming-yang;ZHOU Hai-nan;LI Zeng-hui;LIU Yi-sai;LI Chao;LI Yu(Navigation College,Dalian Maritime University,Dalian 116026,China;Yangtze River Nanjing Waterway Bureau,Nanjing 210011,China)

机构地区：[1]大连海事大学航海学院,辽宁大连116026 [2]长江南京航道局,南京210011

出　　处：《大连海事大学学报》2020年第3期31-37,共7页Journal of Dalian Maritime University

基　　金：中央高校基本科研业务费专项资金资助项目(3132019400)。

摘　　要：为提高内河航道水位预测精度,利用深度神经网络深入研究内河水位的智能预测,提出基于GRU循环神经网络的多站联动水位预测模型.在长江下游多个水位站30年8时水位观测数据集上的实验结果表明,该模型能够综合利用上下游水位站间的水位值关联性,从而减小了单水位站数据随机性的影响,具有较高的预测精确度,其5日预测的最大平均相对误差值MRE优于经典ARIMA模型,也优于单水位站的GRU模型.利用TensorFlow Serving对预测模型进行部署,并通过基于Spring Boot及Vue.js等技术的前后端分离框架开发智能水位预测服务系统,系统的各个部分独立部署,通过RESTful API接口对接,具有很好的松耦合性和灵活性.系统的预测结果可通过Web页面、APP和微信公众号等多种形式展示,为内河航运用户提供了便利的智能服务.In order to improve the accuracy of water level prediction in inland waterway,the intelligent prediction of water level in inland waterway was studied by using depth neural network.A multi station linkage water level prediction model based on GRU cycle neural network was proposed.The experimental results on the water level observation data sets of several water level stations in the lower reaches of the Yangtze River at 8:00 in 30 years show that the model can make full use of the correlation of water level values between the upstream and downstream water level stations,reduce the impact of the randomness of single water level station data,and with a high prediction accuracy.The maximum average relative error value MRE of 5-day prediction is better than the classic ARIMA model,and also better than the GRU model of single water level station type.The prediction model is deployed by using TensorFlow Serving,and the intelligent water level prediction service system is developed by the front-end and back-end separation framework based on Spring Boot,Vue.js and other technologies.Each part of the system is deployed independently and connected through the RESTful API interface,which has good loose coupling and flexibility.The prediction results can be displayed in various forms,such as Web pages,APP and WeChat official account,which provides convenient intelligent services for inland river users.

关 键 词：内河航道 水位预测 门控循环单元(GRU) Tensor Flow Serving 智能服务 

分 类 号：U644[交通运输工程—船舶及航道工程]